Inspired by Advanced Chess, let human players consult a computer in the following limited fashion while playing:
A player starts with and retains the privilege of using the computer only for as long as the computer is being asked to evaluate bad moves, i.e., to confirm that a possible move from a specified position is actually bad. Once the computer evaluates a human-entered proposed move to be good, the player gets no more computer use.
Details remain for specifying what is a "bad move". Maybe more than half a pawn loss in evaluation compared to what the computer thinks is the best move. (Use MultiPV.) Maybe the human can set the threshold.
Can only moves from the current position be evaluated, or moves from arbitrary positions? Probably the latter, as a confirming a future bad move can usefully eliminate a possible line of future play from the current position.
Because a player will typically want to keep computer privilege for as long as possible, the move that causes loss of computer privilege will likely be interesting. The goal is to allow the player to play moves (one move) that otherwise might not have been played: perhaps deemed too risky or outright bad. With the computer's assurance that a risky move "works", the human can steer the game into interesting territory, though the human still needs to find the path through it. Hopefully, this yields more interesting games.
Perhaps the human gets back computer privilege (to then be used and lost again the same way) when both players have lost their (first) computer privilege.
Perhaps several "strikes" before losing computer privilege.
No comments :
Post a Comment